unsplash
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata (descriptions and alt-text) from an external API, which creates a surface for indirect prompt injection attacks.\n
- Ingestion points: Untrusted photo metadata is ingested from api.unsplash.com via the api_request function in scripts/lib/common.sh, utilized by search.sh and random.sh.\n
- Boundary markers: The skill does not implement delimiters or explicit instructions to prompt the agent to ignore potentially malicious text embedded in image descriptions.\n
- Capability inventory: The skill has access to network operations (curl via common.sh) and local command execution via its bash scripts (e.g., search.sh, random.sh).\n
- Sanitization: No sanitization or filtering is applied to the metadata values fetched from the API before they are presented to the agent context.\n- [EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch photo metadata and tracking information from Unsplash (api.unsplash.com), which is a well-known technology service.
Audit Metadata